Failover modes

Client-level failover modes define how backup adapters respond when a primary adapter becomes unavailable. These modes—cold, warm, and hot—range from starting a standby adapter only after failure (cold), to keeping it partially active (warm), to having it fully active and ready to take over instantly (hot), ensuring continuous data flow with varying levels of responsiveness and resource usage.

Failover mode logic

To understand failover strategies, it is useful to examine the three main types—cold, warm, and hot—each varying in recovery speed, resource demands, and complexity:

• Hot

  In hot failover mode, configured components for both the primary and secondary adapters collect and buffer data but only the primary adapter egresses data to the data endpoint. The failover scores for both the primary and secondary adapter are set at 100.

  Data from the secondary adapter instance only egresses the buffered data to the endpoint when a failover event occurs on the primary causing the failover score to drop to zero.

  

• Warm

  In warm failover mode, configured components for the secondary adapter instance start and connect to the data source but do not collect or egress data from the data source. The secondary adapter conducts a test connection check every five minutes to verify that communication with the data source is ongoing. If the connection test passes, the failover score remains at 100. If it fails, the score drops to zero.

  After a failover event occurs, the secondary becomes the primary and begin to collect and egress data to the endpoint.

  

• Cold

  In cold failover mode, the primary adapter connects to the data source and egresses data to the endpoint while you configure the secondary adapter but none of the configured components connect to the data source. The failover score for the secondary adapter is set to 1, meaning that the adapter is in standby mode and will not be promoted to the primary adapter unless the score for the current primary adapter drops to zero.

  After a failover event occurs, the secondary adapter becomes the primary, the failover score is raised to 100, and it begins to collect and egress data to the endpoint.

  

Failover and history recovery

When using client-level failover in combination with the HistoryOnly data collection mode, history recovery does not carry over after a failover event, regardless of which failover mode you use. If the primary adapter fails and the secondary becomes active, it will not resume or complete any previously configured history recovery operations.

If you attempt to configure failover while the RDBMS adapter is in HistoryOnly mode, a warning occurs indicating that on-demand history recovery configurations from the primary adapter will not be completed on the other adapter after failover occurs.

Additionally, if the dataCollectionMode parameter in the Data source configuration is set to CurrentWithBackfill, the adapter behaves as though it is in CurrentOnly mode and history recovery is disabled.